Machine Learning-Driven Prediction of Composite Materials Properties Based on Experimental Testing Data.

Area of Science:

• Materials Science and Engineering
• Computational Materials Science
• Polymer Composites

Background:

• Increasing demand for high-performance, cost-effective composite materials.
• Need for advanced computational methods to optimize composite composition and properties.
• Thermoplastic composites with diverse fillers are crucial for various applications.

Purpose of the Study:

• To apply machine learning (ML) for predicting and optimizing functional properties of thermoplastic composites.
• To investigate the effects of various fibrous, dispersed, and nano-dispersed fillers on composite performance.
• To establish a data-driven framework for rational filler selection in composite design.

Main Methods:

• Material synthesis via powder metallurgy.
• Microstructural analysis, mechanical testing, and tribological testing.
• Development and validation of ML regression models for property prediction (R-squared up to 0.80).

Main Results:

• Optimal filler selection significantly enhances wear resistance (e.g., carbon fibers 17-25x, kaolin 45-57x) while managing mechanical strength.
• Specific fillers like basalt fibers, kaolin, coke, graphite, sodium chloride, titanium dioxide, and PTFE showed distinct property enhancements.
• ML models effectively predicted composite properties, explaining up to 80% of variability, reducing extensive experimental needs.

Conclusions:

• Machine learning provides an efficient, cost-effective framework for optimizing composite materials.
• The study demonstrates the potential of ML in guiding filler selection for tailored composite properties.
• This approach contributes to sustainable industrial practices and aligns with Sustainable Development Goal 9.